Replay Simulations with Personalized Metabolic Model for Treatment Design and Evaluation in Type 1 Diabetes

Area of Science:

• Biomedical Engineering
• Computational Biology
• Endocrinology

Background:

• Personalized assessment of type 1 diabetes mellitus (T1DM) treatment strategies requires replaying real-life data.
• Previous methods for replaying T1DM data had a limited domain of validity.
• Extending the applicability of data replay methods can lead to true personalization of glucose control.

Purpose of the Study:

• To propose and test an enhanced method for replaying type 1 diabetes mellitus (T1DM) data in silico.
• To extend the applicability of subject-specific model personalization for glucose control.
• To improve the accuracy of simulating "what-if" scenarios for T1DM treatment.

Main Methods:

• Subject-specific model personalization of insulin sensitivity and meal absorption parameters.
• Utilizing the University of Virginia (UVa)/Padova T1DM simulator for generating and testing scenarios.
• Assessing method performance using Mean Absolute Relative Difference (MARD) and Clarke Error Grid Analysis (CEGA).

Main Results:

• Model personalization decreased MARD by 9.08% for basal and 6.07% for bolus insulin changes compared to prior methods.
• Replay simulations demonstrated high accuracy, with MARD <10%.
• Over 95% of simulated glucose readings fell within the CEGA A-B zones, indicating clinical accuracy.

Conclusions:

• The proposed method for replay simulation is numerically and clinically valid.
• The method demonstrates robustness over a wide range of scenario input changes.
• This approach shows potential for optimizing T1DM treatment strategies.